technique

L2 perturbation

techniqueactivel2-perturbation-2a03e29b·1 events·first seen 28d ago

Aliases: L2 perturbation

Co-occurring entities

adversarial robustness L-infinity perturbation adversarial training OpenAI

More like this (12)

L-infinity perturbation LPU LLM-Guided Model Patches hallucination (LLM)RL² LLM inference KL Divergence lesioning technique Latent Adversarial Robustification (LAR)VulnLMP TensorRT-LLM L0 regularization

Recent events (1)

4Openai Blog·28d ago·source ↗

Transfer of Adversarial Robustness Between Perturbation Types

OpenAI published research examining whether adversarial robustness trained against one type of perturbation (e.g., L-infinity) transfers to other perturbation types (e.g., L2, L1). The work investigates the generalization properties of adversarial training across different threat models. This is an early safety and robustness research contribution from OpenAI predating the modern LLM era.

Evaluation and Benchmarking AI Safety Research adversarial robustness L-infinity perturbation adversarial training +2 more